How Do I Get ChatGPT to Recommend My Consulting Firm?

You get recommended by winning the retrieval path, not the memory path. When a buyer asks ChatGPT for help in your space, the names in the answer come from one of two places: what the model absorbed in training, which favors the biggest and oldest firms in your category, or a live search the assistant runs before it writes, which favors pages and third-party sources it can find and parse right now. A twelve-person advisory shop will almost never beat a global consultancy on the first path. It can absolutely beat one on the second.

Seven moves do that work: write down the exact prompts your clients type, measure what all four assistants say today, stop counting rank and start counting mentions and cited sources, make your existing pages retrievable, answer the question in your first paragraph, get described the same way on domains you do not own, then re-ask the identical prompts and follow the money.

Step 1: Write down the exact prompts a buyer would type

Not keywords. Sentences. "Who can help us fix revenue attribution before our Series B" is a prompt. "Marketing consultant" is not.

Include the two modifiers most firms leave off: specialty and place. "Best fractional CFO for early-stage SaaS in Raleigh in 2026" reads differently to an assistant than "best CFO consultant." Recency and location cues make an assistant more likely to go search before it answers, and searching is the path where a small firm has a shot. Ten prompts is enough to start, and you will learn more from ten you actually track every week than from fifty you write once and forget.

Step 2: Check all four engines, and check them more than once

ChatGPT, Claude, Gemini, and Perplexity assemble answers differently. Perplexity cites more sources and skews recent. ChatGPT with browsing cites fewer and leans on domains it treats as authoritative. Gemini leans on Google's index signals. Checking one engine once tells you almost nothing, and personalization plus chat memory mean the answer in your own logged-in account is not the answer your buyer gets.

That is the flaw in the manual spot-check every owner starts with: you type your prompt, feel a jolt of dread or relief, and act on a sample size of one. Clarity's free AI visibility check gives you a read across the engines without your own history skewing it. On paid plans Clark runs every day, prefixing each ask with a buyer persona and a location line so the answers reflect what a prospect would actually see, and rotating through your questions so each one comes back around roughly weekly.

Step 3: Drop the word "rank"

There is no rank inside an AI answer. Three numbers matter: what share of answers mention you (mention rate), where in the list you land when you are named (average position), and which sources the engine cited to build the answer. That third one is your map. We deliberately do not roll these into a single "AI visibility score," because a made-up composite number tells you nothing about what to fix. If you want the longer argument for measuring this way, we made it in our comparison of closed-loop attribution against rank-only tracking.

Step 4: Make your existing pages machine-readable before you write more of them

Most consulting sites lose here quietly. Four things carry the weight: whether AI crawlers are allowed in at all, whether your content is structured so a retrieval layer can lift a clean passage, whether your schema markup describes your services and location in a way engines can trust, and whether you publish an llms.txt map of your site. Run a free AI readiness check and fix the failures before you commission a single new page. A brilliant service page behind a blocked crawler is invisible.

Step 5: Answer the question on the page, in the first paragraph

The biggest structural lever is answering first. Open with a direct, self-contained answer to the exact question, because that is the kind of passage an engine can quote without stitching anything together. Shape your headings like the prompt. Name your method, your client situations, and your real numbers. A hedged, generic paragraph loses to a specific, confident one every time, which is most of how a post actually earns a citation.

Step 6: Get described the same way somewhere you do not own

This is the part owners resist. Assistants weigh agreement across independent sources heavily, and for "best firm in my category" questions much of what they cite is not vendor-owned at all: directories, review sites, roundups, association pages, forum threads. Twenty more pages on your own domain does not create agreement. Being described consistently across five independent domains does.

So pull the source URLs the engines actually cited for your prompts, and treat the recurring ones as a target list. Ask for the same three-line description of what you do and who you do it for on every one, so an assistant reading two of them sees the same firm. That off-site work is yours to do, and no tool does it for you, ours included.

Step 7: Re-ask the same prompt, then follow the money

Reuse the question wording word for word so before and after are comparable. Then check whether the traffic converts. AI referrals arrive as a small slice of sessions, and much of it lands in analytics looking like direct traffic, so most firms undercount the channel badly. Semrush found LLM-referred visitors convert 4.4x higher than baseline web traffic, because the assistant already pre-qualified them.

Unmentioned is not a permanent condition. One leather shoe brand, Bosten Shoes, went from absent to the top recommendation in its category. A funded B2B SaaS team we work with landed its first ChatGPT-attributed sale in under 30 days, and it showed up as named revenue rather than another line of anonymous direct traffic. Everyone can show you where you rank. Proving the mention paid is the harder job, and the one worth building toward. Start with the two free checks above, then decide whether you want the loop running itself.